The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear Wenqing Zhang,

Wenqing ZHANG wrote:
> Dear Prof. Gonze,
>
> When I did calculation for BaTiO3, I met a convergence
> problem. No matter iscf = 3 or 5 I used, it was always
> very hard to get tolwfr less than 1.d-9. After nearly
> 20 times iteration, this quantity begins to fluctuate
> around 2.d-8. 

BaTiO3 was already studied by Philippe Ghosez using ABINIT, many years
ago, without any problem, so your result is strange. He is still
studying it, for different purposes, so you might contact him.

In any case, the wavefunction residual might not be a very good indicator
of the real SCF convergence. If
you are taking the default number of bands,
some conduction bands are included. States explicitely
treated might be degenerate
at some k-points with states NOT treated, which screw up the
convergence, but ONLY for these degenerate states.
Such a degeneracy is often responsible of the
lack of convergence of the residual on the wavefunctions
(that is the maximum residual on ALL wavefunctions, including
the unoccupied ones).

Good continuation,
Xavier



Is there any good way to speed up the
> convergence of iteration?
>
> I used broden mixing before. It is very stable as I
> remembered. Does ABINIT have broden mixing scheme?
>
> Attached is my input file. Please let me if there is
> any problem. Thanks for help.
>
> Yours -- Wenqing
> ----
> ----- Copy of BaTiO3.in ---
>
> #baTiO3.in
> #DATASET 1 : SC run
>   prtden1  1
>    ngkpt1  2 2 2     # k mesh
>   kptopt1  1     # Automatic K mesh with full symmetry
>   tolwfr1  1.0d-9
>   getwfk1  1        # read in WFK of dataset1
> #  getden1  1        # read in DEN of dataset1
>
>
> #Common data
>  shiftk  0.5 0.5 0.5
>
>  acell   3*7.57
>  rprim   1.0 0.0 0.0
>          0.0 1.0 0.0
>          0.0 0.0 1.0
>  natom   5
>  ntypat  3
>  typat   1 2 3 3 3       # Ba, Ti, O, O, O
>   xred   .0000   .0000   .0000
>          .5000   .5000   .5000
>          .5000   .5000   .0000
>          .5000   .0000   .5000
>          .0000   .5000   .5000
>  znucl    56.0     22.0    8.0     # Ba, Ti, O
>   amu    137.34003 47.88  16.00    # mass of atoms
>
>  iscf    3   # 5 too
>  ixc     3
>  diemac  6.75
>  ecut    45.00     # hartree, around 45.0
>
> # occopt  1        # = 1: no occ array to be input
> here, and band is just one number (elec#/2-unspin)
> # nband   20       # 21 with an empty band also tried.
> # nbdbuf  0
>  nstep   250
>
>
> --- potential used --
>
> ../../Psps_for_tests/56ba.psp.mod
> ../../Psps_for_tests/22ti.psp.mod
> ../../Psps_for_tests/8o.psp.mod
>
>
>
>
>
>
>
>
>
>
>
> =====
> Wenqing Zhang
> 1970 Crystal Lake Dr. Apt#E79
> Shelby Twp., MI 48316
> Tel: (586)254-9931(h), (586)323-1551(o)
>
> __________________________________
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site design software
> http://sitebuilder.yahoo.com